Node label prediction using relational classifier and checking the assumption of homophily is reasonable for given network? Divide the node label into $3$ part: Training, Validation and Test set using K$-$fold cross validation set.

i$.$ Use training set for training the model

ii$.$ Use Validation set for optimization of any hyper parameter you have in your model.

iii. Use Test set only for calculating accuracy of the model $($ Confusion matrix generation based on this$)$

c$.$ Develop Relational classifier as explained in attached document

d$.$ Check the accuracy of unlabeled node and produce confusion matrix for this